My Journey Through "Where Good Ideas Come From"

Steven Johnson dismantles the myth of the lone genius having a eureka moment. Good ideas don't work that way. They emerge slowly, at the edges of what's already known, from collision and recombination.

The Adjacent Possible

Johnson's best concept. Innovation happens at the boundary of current capabilities. You can only open doors that are adjacent to where you already stand. Each breakthrough unlocks new possibilities that weren't available before. This maps perfectly to how I've experienced software architecture decisions — you build on what exists, and the best solutions come from recombining existing pieces in new ways.

The Slow Hunch

Breakthrough ideas rarely arrive fully formed. They incubate. Weeks. Months. Sometimes years. Johnson calls this the "slow hunch." I've experienced this directly. A problem sits in the back of my mind, and one day in the shower or on a walk, the solution clicks. It feels sudden. It isn't. The background processing was running the whole time.

Liquid Networks

Ideas need collisions. Johnson argues that the best environments for innovation are "liquid" — not too rigid, not too chaotic. Coffee houses in the Enlightenment. Open-plan research labs. Slack channels where unrelated teams overlap. Connection density drives innovation.

What I Took Away

I started keeping a "spark file" after reading this. Half-baked ideas, random observations, things that seem interesting but don't have a home yet. Johnson convinced me that the raw material of innovation is just sitting around waiting to be connected.

Where It Falls Short

Johnson's biological metaphors — coral reefs, ecosystems — sometimes stretch too thin. Not every innovation pattern maps cleanly to nature. And the book occasionally reads like a TED talk extended to book length. The core ideas are strong, but some chapters repeat the thesis rather than advancing it.

Still, this fundamentally changed how I approach problem-solving. Less "think harder." More "connect wider." That's a better model.